Geant versions: 4.8.2p01 4.9.0 4.9.1 4.9.1p01 OS:Linux Debian/testing. Behavior of NeutronInelastic changed since Geant4.8.2p01. On an older versions (4.9.*) it gives another result. The example application is attached (it needs ROOT installed). It simulates neutron propagation in cube, filled with oxygen. It simply saves delta Energy for neutron on each step and plot histogram for energy loss due to inelastic scattering. The effect is seen only when initial neutron energy is more than 20 MeV (21 in example). Energy loss per step distribution have a large peak on low energy for g4.8.2p01, when g4.9* doesn't have it. Step-by-step: - Compile example with G4.8.2 and with G4.9.1. - Run them separately with test.mac (be sure that initial energy is 21 MeV) - Compare result files with root macros compare.C - Look at energy loss per step distribution. I'm not sure if it is an error, but I can not find anything about it in Release Notes. Thank you. regards, Maxim Gonchar.
Created attachment 21 [details] example application
Created attachment 25 [details] script for comparison Script that compares output of example application. Sorry, I forgot to attach it.
The correct behavior is that there should not be a large peak at low energy. This problem was actually fixed in Geant4 9.0, according to the 9.0 Release Notes. Please see under Hadronic Physics Bug fixes in LEP models.
